Liquid AI introduces LFM3

Liquid AI has introduced LFM3, a family of small models for text, vision, audio and decision tasks. The company says the models share a common backbone and are open-weight, so businesses can fine-tune them on their own data and run them locally, including on edge devices. The first release includes six models around the 2B-parameter range. Liquid AI says its text model fits in about 1 GB at 4-bit precision and has a 256K-token context window. Those are company-reported specifications, not a guarantee that every model will suit every device or task.

For an operator, the practical idea is to choose a model for a defined job, adapt it, and run it where the application is used, rather than relying on one large general-purpose model for everything. That could matter when a workflow needs to keep data on a local system or work within a device’s limits. The trade-off is that local deployment still needs testing: a compact model may be easier to run, but it must handle your actual business task reliably. Liquid AI says the models can be fine-tuned on a single GPU and provides instructions for local use in its LFM3 announcement. This is the kind of capability we build into an AI command centre, where teams can keep track of models and the work they’re suited to.